On Thursday 04 July 2002 09:51, you wrote: > On Thu, 4 Jul 2002 03:50:29 +0200 > > Dieter N�tzel <[EMAIL PROTECTED]> wrote: > > On Thursday 27 June 2002 11:37, you wrote: > > > On Tue, 25 Jun 2002 22:47:23 +0200
> > > what settings do you have in the BIOS? > > > > > > you get about 100MB/sec more than me (~10%) > > > > Memtest 3.0 measured: > > ~600 MB/s for the former 512 MB CL 2 module > > ~686 MB/s for 1 GB (2 x 512 MB CL 2.5 in CL 2 mode) > > where can I get memtest? http://www.memtest86.com/ > > First some Athlon memory numbers: > > > > SunWave1 nuetzel/Entwicklung# ./athlon > > is that part of memtest? No, but I've sent them to you. They are from times when the MMX/3DNow! memcopy routines are made ;-) Find both source files in the attachment. > > Four runs with 2.4.19rc1aa1, today. > > hm. maybe thats the difference? (2.4.19-pre8-ac here) Maybe, but not that much, I think. But I have an experimental page coloring patch (a module) from time to time running, too. But it wasn't loaded during my AGP tests. I can make a "unified" patch on top of 2.4.19-rc1 if you like. > > SunWave1 Entwicklung/Kernel# ./testgart > > Average speed: 1418 mb/s > > sweet. Yes ;-) > > MPS Version control: 1.4 (1.1 works, too) > > Video BIOS Shadow: disabled > > System BIOS Cacheable: disabled > > Video RAM Cacheable: disabled > > why the above disabled?? As I read on several sources no current OS (Linux, *BSD) make use of them 'cause they access "directly" the hardware. But I'll recheck even if Video RAM Cacheable make some difference. > > BIOS of MP Capabilities: enabled (disabled works, too) > > Memory Hole At 15M-16M: disabled (as always ;-) > > AGP Aperture Size (MB): 64 (e.g. my current V5 5500 do not use it) > > if the V5 doesn't use it, how can you test it? We are currently testing "only" the pure AGP transfer rate not the card's ability. Or is this the starting point four a much faster tdfx driver??? I had much better numbers with the beta version of Xig's Accelerated-X server for some benchmarks. They had a much better AGP GART driver at that time. Please have a look at the "geometry and fill rate": >>> My dual Athlon MP 1900+ system <<< Mesa/demos> ./tenmillion smooth running on SunWave1 (Mesa DRI 20020221 Voodoo4 x86/MMX/3DNow!) Thu Jul 4 16:35:47 2002 visual: 0x23 50 pixel triangles, 90 triangles/strip, 1 strips vertex data: v3f smooth shading, zbuffer off, light off, texture off, immediate off test time: 0.768 secs. geometry rate: 928920 triangles/sec fill rate: 46.446 Mpixels/sec draw_strips_count: 1 Mesa/demos> ./tenmillion area=3 smooth running on SunWave1 (Mesa DRI 20020221 Voodoo4 x86/MMX/3DNow!) Thu Jul 4 16:33:42 2002 visual: 0x23 3 pixel triangles, 90 triangles/strip, 4 strips vertex data: v3f smooth shading, zbuffer off, light off, texture off, immediate off test time: 2.000 secs. geometry rate: 4591122 triangles/sec fill rate: 13.773 Mpixels/sec draw_strips_count: 1 >>> Now the older 1 GHz Athlon II SlotA with Accelerated-X <<< SunWave1>./tenmillion area=3 running on SunWave1 (VSA-100/3DNow!) Thu Sep 27 05:25:28 2001 visual: 0x21 3 pixel triangles, 90 triangles/strip, 4 strips vertex data: v3f flat shading, zbuffer off, light off, texture off, immediate off test time: 1.996 secs. geometry rate: 6836102 triangles/sec fill rate: 20.508 Mpixels/sec draw_strips_count: 1 SunWave1>./tenmillion area=3 smooth running on SunWave1 (VSA-100/3DNow!) Thu Sep 27 05:25:39 2001 visual: 0x21 3 pixel triangles, 90 triangles/strip, 4 strips vertex data: v3f smooth shading, zbuffer off, light off, texture off, immediate off test time: 1.996 secs. geometry rate: 6836527 triangles/sec fill rate: 20.510 Mpixels/sec draw_strips_count: 1 > > AGP Data Transfer Mode: 4x (auto detection lower it to 2x according V5) > > you get this speed at 2x ? Yes. The 3dfx V4/5 support only AGP 2x. 01:05.0 VGA compatible controller: 3Dfx Interactive, Inc. Voodoo 4 / Voodoo 5 (rev 01) (prog-if 00 [VGA]) Subsystem: 3Dfx Interactive, Inc.: Unknown device 0002 Control: I/O+ Mem+ BusMaster- SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B- Status: Cap+ 66Mhz+ UDF- FastB2B+ ParErr- DEVSEL=fast >TAbort- <TAbort- <MAbort- >SERR- <PERR+ Interrupt: pin A routed to IRQ 17 Region 0: Memory at e0000000 (32-bit, non-prefetchable) [size=64M] Region 1: Memory at d8000000 (32-bit, prefetchable) [size=128M] Region 2: I/O ports at b000 [size=256] Expansion ROM at <unassigned> [disabled] [size=64K] Capabilities: [54] AGP version 2.0 Status: RQ=15 SBA+ 64bit+ FW- Rate=x1,x2 Command: RQ=0 SBA- AGP- 64bit- FW- Rate=<none> Capabilities: [60] Power Management version 2 Flags: PMEClk- DSI+ D1- D2- AuxCurrent=0mA PME(D0-,D1-,D2-,D3hot-,D3cold-) Status: D0 PME-Enable- DSel=0 DScale=0 PME- > I am running at 4x and am not faster! > > > AGP Always Compensate: enabled > > AGP Secondary Lat Timer: 20h > > what is the above? Don't know exactly...;-) But I have a very good connection to MSI (Germany) and can ask some one there. > > Idle Limit: 12 Cycles > > Page Hit Limit: 16 > > Trc Cycle: 5 > > Trp Cycle: 2 > > Tras Cycle: 2 > > CAS: 2 > > Trcd Cycle: 2 > > looks like my memory timings too ;) > > > Is this enough? > > Its a start ;-) Nah, then let's move 8-) -Dieter -- Dieter N�tzel Graduate Student, Computer Science University of Hamburg Department of Computer Science @home: [EMAIL PROTECTED] ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Caffeinated soap. No kidding. http://thinkgeek.com/sf _______________________________________________ Dri-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/dri-devel
